Abstract: In this study, over one million medical record data were collected from consultations and some knowledge graph algorithms were utilized to integrate clinical information, imaging, diagnosis and other heterogeneous data sources. Linking disease, symptoms, medication, and gene information based on external datasets, a medical knowledge graph was constructed involving over 6800 entity and 330000 relationship data of the field of neurosurgery. Fifteen sets of high-frequency relationship data were validated according to the relationship between diseases, drugs, and symptoms. Enrichment analysis was used to explore the correlation between various entities, and the Neuro4j was applied to visualize the knowledge graph of neurosurgery. Based on knowledge graph algorithms, the missing information of clinical and microscopic in diagnosis process was provided as the accurate and detailed diagnostic basis for clinicians to improve diagnostic accuracy and treatment efficiency. Analysis results show that based on the established knowledge graph, knowledge exploration can be achieved to assist clinicians in discovering new disease features, pathogenesis, drug targets, et al, promoting knowledge innovation in the field of neurosurgery medicine.